Dittos, 

ADVERTISEMENT
ADVERTISEMENT
Back Home Button
The Rush Limbaugh Show
Excellence in Broadcasting
RSS Icon
ADVERTISEMENT

EIB WEB PAGE DISGRONIFIER

Caller's Analysis of Skewed Poll Samples

BEGIN TRANSCRIPT

RUSH: Here is Ed in Fort Myers, Florida, as we head back to the phones.  Ed, great to have you on the EIB Network.  Hi.

CALLER:  Well, thank you, Rush.  It's a thrill to talk to you.

RUSH:  Thank you, sir.

CALLER:  I hate to think where we would be if it weren't for you.

RUSH:  I think the same thing, I must tell you.

CALLER:  Ha-ha.

RUSH:  Ha-ha-ha-ha.

CALLER:  My comment concerns the oversampling in these polls and the way that's reported or described.  If there were a poll that is 30% Republican, 40% Democrat, the description is a 10% oversampling of Democrats.  But that's not really correct.  If it were a hundred people in the poll, it would 30 Republicans and 40 Democrats, that would be 10 Democrats more than Republicans, and that would be 10 out of 30, which is a 33% oversampling of Democrats.

RUSH:  Oh, you're gonna go all engineer on us here?

CALLER:  Yeah, he wrote that down, huh?

RUSH:  All right.  Numbers are really hard for people when they can't see them, even me.  You got 40% Democrat, 30% Republican, that's reported as a plus 10 Democrat advantage, but you say it's actually plus 30, in terms of percent?

CALLER:  Thirty-three.

RUSH:  Thirty-three.  Okay, how do you get there?

CALLER:  Well, if there were a hundred people in the poll, there would be 30 Republicans and 40 Democrats.

RUSH:  Right.

CALLER:  That would be 10 more Democrats than Republicans.

RUSH:  That's right.

CALLER:  And that would be 10 out of 30, because there were 30 Republicans, and that's 33% more Democrats than Republicans.

RUSH:  Why do you take 10 out of 30 and not 10 out of a hundred?

CALLER:  Because there are only 30 Republicans.  If they sample 30 Republicans and 30 Democrats, it would be zero.  But they sampled 10 more Democrats than they did Republicans and there were only 30 Republicans in the first place.  So that would be 33% more Democrats than Republicans.

RUSH:  Well, there's no question the math works the way you're doing it.  The only question that I anticipate the audience having is, even though there are only 30 Republicans in the poll, we still have a sample of a hundred people here.  We've got 40 Democrats, 30 Republicans, that leaves 30 -- we'll say independents, undecideds, which you're throwing out, they don't matter for what you're doing, right?

CALLER:  No, they don't matter because we're only talking about the difference between Republicans and Democrats.  The difference being 10 out of 30.

RUSH:  Okay.  All right, then here's the next question.  If what you say is right, 33% more Democrats than Republicans --

CALLER:  Yes.

RUSH:  -- then how does that manifest itself in the results that the poll says?

CALLER:  Well, the poll is skewed by 33%.

RUSH:  Okay, so a poll that shows -- let's get one.  And bear with me here.  I'm way behind you on numbers.  I'm not good with numbers.  Forty Democrats, 30 Republicans, and let's say the poll shows Obama 49, Romney 47.

CALLER:  Yeah.  Wait a minute, that's not the poll.  That's the poll result, yes.

RUSH:  That's what they're reporting.  The poll of 49-47 with a sample of 10 more Democrats than Republicans. 

CALLER:  Well, it depends on how many they sampled, yes.

RUSH:  Right.  Well, I'm using your number, 40 and 30.

CALLER:  Yes.

RUSH:  So I've got 40-30, and then 30.  So got 40 Democrats, 30 Republicans, you say there's 33% more Democrats.  Okay, that poll -- this is hypothetical -- shows Obama 49, Romney 47.  Does that mean, according to your calculations, that we have to take 30% away from Obama, away from 49%?  What I'm getting is, how does this --

CALLER:  I haven't actually thought that through, but I think it would be something like, if you're gonna compare only the results in the percentages of 49-47, then it would be 10 out of 60.  In other words, if you did a proper poll, there would be 30 Republicans and 30 Democrats.  We don't have 30 Democrats.  We have 40 Democrats, which is 10 more out of the base of 30 for Republicans.  If you say 33% more Democrats than Republicans --

RUSH:  Yeah, but what I'm trying to get --

CALLER:  -- so it would be 10 divided by 70 as a percentage spread on the results of the poll.  I know I lost you there.

RUSH:  Well, only because I'd have to divide it on paper to get what that number is.

CALLER:  It's real simple.  I sample 30 Republicans, and I sample 40 Democrats.  How many more Democrats did I sample?  Ten.  Right?

RUSH:  No, no.  I understand all this.  What I'm trying to get to is what does it mean when they also tell us that of these 40 and 30, that 49% prefer Obama and 47 prefer Romney, what does the 33% -- does it change --

CALLER:  Oh, yeah.

RUSH:  Well, how? 

CALLER:  It's going to change it in terms of a higher percentage for Romney in that case.

RUSH:  Yeah, and the way people are doing it, in your example, if we got a 49-47 Obama lead with a sample of 10 more Democrats, then we're thinking it should be --

CALLER:  Probably 2%.

RUSH:  Two percent.

CALLER:  Yeah.

RUSH:  Okay.  Out of the 49-47 it's really 51-46.

RUSH:  For your call to mean something here, we've gotta know what your 33% calculation means to the final result that they are reporting --

CALLER:  It's going to skew it substantially.  In other words, the 49-47 example that you picked would probably be somewhere around 51-46 in reality.

RUSH:  For who?

CALLER:  The 49 being Obama.  Excuse me.  It would be the other way around.  Sorry.  I got that backwards.  If they sample an equal number of Republicans and Democrats, the percentages would -- the improvement in the poll would be beneficial to the Republicans.

RUSH:  Well, there's that 30% undecided we're not factoring here.

CALLER:  No, because --

RUSH:  The only reason I'm asking you, you're trying to conclude or you're trying to get people to conclude that the polls are even more phony --

CALLER:  Oh, yes.

RUSH:  But I don't see how.  I don't know what to make of the 49-47 --

CALLER:  Well, if you had, you know, you do your own poll.  You sample 30 Republicans, you sample 40 Democrats.  Why did you sample 40 Democrats?  You sampled --

RUSH:  Well, ostensibly because --

CALLER:  You sample 10 more Democrats than you did Republicans.  Right?

RUSH:  Right.

CALLER:  So that's a 33% more Democrat than Republicans.  Than Republicans.  It's not 33% of the total.  It's 33% more than they did Republicans.

RUSH:  Right.  No, I understand all that.  See, if you have Obama at 49, Romney at 47 with a Democrat plus 10 sample, then people will conclude that Romney might be ahead by 5 rather than down by 2.  And they're trying to get that.  They're just using the 10 greater.  When you change the way this is looked at to 33% more Democrats, then people are gonna conclude that Romney's even further ahead, or better, that Obama's even further behind, and I just don't know that that works.  Anyway, I appreciate the call and I think statistically you are correct, that that's the way to analyze the number differential, but I don't know how anybody can take that and then apply it to what the reported end-of-poll data is and make the adjustment.

BREAK TRANSCRIPT

RUSH: Okay, before we move on, let me explain what the previous caller was just trying to say. It's a semantic argument only. In the example he gave where there are 40 Democrats and 30 Republicans sampled, the conventional wisdom now -- the way of reporting that -- is a Democrat-plus-10 sample, and in raw numbers, it is. His point was that it's much greater than 10 in reality, that it is a 33% increase in the number of Democrats sampled, not 10.

That doesn't change the outcome. It's simply his way of driving home how out of proportion the oversampled Democrat polls are. Nothing more than that. You have 40 Democrats and 30 Republicans, and it's reported as a Democrat plus 10. It is in raw numbers. But in terms of the percentage, that polling unit actually talked to 33% more Democrats than Republicans, which would help to illustrate just how out of kilter such a poll is with such a sample.

When you won't find a sample like that in election returns. You will not find Democrats plus 10, say in 2010, in terms of people showed up. Turnout, if you will. So it's just a semantic thing. It doesn't change the outcome of the poll as reported. It's just a more accurate way of looking at the out-of-balance situation among those polled.

END TRANSCRIPT

ADVERTISEMENT

Rush 24/7 Audio/Video

Listen to the Latest Show Watch the Latest Show

original

Facebook

ADVERTISEMENT

Most Popular

EIB Features

ADVERTISEMENT: